JPH08320697A - Speech recognition device - Google Patents

Abstract

PURPOSE: To provide a speech recognition device which is suitable for the improvement of recognition rate and shortening of recognizing time and allows the use of partial dictionaries also for the objects such as operation maintenance of general plants or car navigation by changing the partial dictionaries beforehand divided in response to the change of circumstances. CONSTITUTION: The device recognizes inputted voices by beforehand dividing a speech recognition device having voice pattern into partial dictionaries and changing the partial dictionaries to recognize input voices. And it has a position detecting means 1001 which takes the position of a voice input person or the present position of the car, a partial dictionary selecting means 1201 which changes the partial dictionary based on the voice input person or the present car position and a speech recognition means 14 which recognizes the inputted word by comparing the selected partial dictionary with the inputted voice pattern.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition device used for inspection / maintenance of a plant, navigation of a car and the like.

[0002]

2. Description of the Related Art Recently, use of a voice recognition device in a man-machine system in an operation control room of a power plant has been promoted to further improve the recognition rate (prevention of erroneous input) and shorten the recognition time (operation). Improvement) is desired.
Current speech recognition devices perform input speech recognition processing by matching the input speech with the speech pattern of the vocabulary registered in the speech recognition dictionary, so it is necessary to register all the vocabulary to be input in the speech recognition dictionary in advance. is there. It is generally known that the recognition rate decreases when the number of registered vocabularies in the voice recognition dictionary is large and / or the number of similar vocabularies is large. Even if a voice input device is used to inspect the power plant, the system and / or
Alternatively, since the names of devices are enormous, it is necessary to improve the recognition rate and shorten the recognition time. Therefore, a method has been devised in which the vocabulary is divided into partial dictionaries and registered, and the dictionaries are switched to reduce the target vocabulary. For example, in Japanese Patent Laid-Open No. 61-35498, a partial dictionary is created in advance for each type of word as a voice recognition dictionary, and the order of words to be voice input (for example, 1. sex, 2. surname, 3. name, 4. Prefecture name ...) is defined, and the dictionary is sequentially switched according to the input of words in this order. When trying to use the technology of this publication for plant inspection,
A partial dictionary is created for each system and / or each device, and the inspection order and inspection time are specified in advance, and the partial dictionaries are switched. However, in the plant inspection work, the inspection items and inspection routes may be changed, and the inspection time also changes depending on the situation such as the occurrence of an abnormality and the progress of the work. It is impossible to select a partial dictionary by doing. Similarly, in a car navigation system, the travel route may be changed, and the time required to reach the destination may change due to traffic congestion on the travel route. It is impossible to select a partial dictionary by.

[0003]

As described above, in the conventional speech recognition apparatus, it is possible to divide the recognition dictionary into subdictionaries in advance and switch the recognition dictionaries in a prescribed order to use them. Since it cannot be switched correspondingly, it was not applicable to plant inspection work and car navigation systems.

The present invention has been made in consideration of such a situation, and its purpose is to operate a general plant by switching the partial dictionaries divided in advance in response to the change of the situation. It is an object of the present invention to provide a voice recognition device which makes it possible to use a partial dictionary even for an object such as maintenance or car navigation, and which is suitable for improving the recognition rate and shortening the recognition time.

[0005]

The above object can be achieved by detecting the position of a voice input person or the current position of a vehicle and switching the partial dictionary based on the detected position.

[0006]

According to the present invention, the voice recognition dictionary is created in advance as a partial dictionary divided for each system and / or device of the plant, the position of the voice input person is detected, and the system for the position where the voice input is made and / or Alternatively, the partial dictionary is selected according to the change in the inspection situation by identifying the device installed near the position where the voice input is performed. This allows
It is possible to improve the recognition rate and shorten the recognition time of the voice recognition device used in the plant inspection system. The present invention also creates a voice recognition dictionary as a partial dictionary that is divided in advance by city group name, detects the current position (latitude and longitude) of the vehicle, and refers to the map data while referring to the city group name of the current position. Is specified, the partial dictionary is selected according to the change in the driving situation. As a result, it is possible to improve the recognition rate and shorten the recognition time of the voice recognition device used in the car navigation system.

[0007]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows a voice recognition device of a plant inspection system according to a first embodiment of the present invention. The voice recognition device includes a position detection unit 1001 that detects the position of a voice input person, a plant information storage unit 1101 referenced by the position detection unit 1001,
Partial dictionary selection unit 1201, voice recognition dictionary (partial dictionary) storage unit 1301, voice recognition unit 1401, computer 1501,
Inspection record storage 1502, CR for displaying inspection records
T1503, printer 1504 for outputting inspection report based on inspection record, transmitting / receiving unit (fixed station) for transmitting data
1601, a transmission / reception unit (mobile station) 1602, a mobile display 1701, and a microphone 1801.

FIG. 2 is a schematic diagram of a plant site used for explaining the first embodiment. In FIG. 2, in the plant site, a turbine system, a circulating water system in which a circulating water pump A, an auxiliary pump and a circulating water pump B are arranged, an oil filter, T / D-RFP-A, an N36 valve and a valve LPSV are arranged. It consists of a water supply system, an on-site power supply system where a generator and a hydrogen gas pressure gauge are arranged, and a main steam system. Each system is partitioned by a wall, and gates are provided to allow access to each system. Sensors (S1 to S5) 1002-1 for detecting that a maintenance person (voice input person) who inspects equipment installed in each system has passed through each gate
006 is set. Routes for inspecting equipment installed in each system are defined, and route 1 and route 2 are shown in FIG. The route 1 and the route 2 are selected in advance according to the day or the time (for example, morning and afternoon), and the voice input person basically performs the inspection work along this route. For example, in route 1, the voice input person enters the circulating water system through the gate after inspecting the turbine system equipment. At this time, the sensor (S1) 1002 detects the passage of the voice input person.
Next, the voice input person inspects the circulating water pump A, the auxiliary pump, and the circulating water pump B arranged in the circulating water system, and enters the water supply system through the gate. At this time the sensor (S2)
1003 detects passage of a voice input person. And so on
Check each equipment installed in the water supply system and the power supply system in the plant and enter the main steam system. At that time, the sensor (S3) 1004,
A sensor (S5) 1006 detects passage of a voice input person. The same applies to the route 2. Here, the sensors (S1 to S5) 1002 to 1006 use a semiconductor laser and a photodetector, and are installed so that the laser light always enters the light receiving surface of the photodetector. When the voice input person passes between the semiconductor laser and the photodetector, the sensor detects the passage of the voice input person by blocking the laser light incident on the light receiving surface of the photodetector.

FIG. 3 shows the sensor installation position data stored in advance in the plant information storage unit 1101 referenced by the position detection unit 1001. The sensor installation position data represents between which systems each sensor is installed. As is clear from FIG. 3, the sensor (S1) is installed at the position of the gate partitioned by the walls of the turbine system and the circulating water system. The same applies to the other sensors (S2 to S5).

FIG. 4 shows inspection route data stored in advance in the plant information storage unit 1101 referenced by the position detection unit 1001. In the inspection route data, inspection routes are stored in order by system name. Inspection number 1 (route 1) indicates that the inspection is performed in the order of turbine system-circulating water system-water supply system-in-house power supply system-main steam system. Inspection number 2 (route 2)
The same is true for

FIG. 5 shows the position detector 1 according to the first embodiment.
The processing of 001 is shown. When the process is activated, it is checked whether the sensor detects a voice input person (step 30).
01), the sensor that has detected the voice input person is specified (step 3002). Next, the setting position data of the sensor that has detected the voice input person is read (step 3003),
The inspection route data is read (step 300
4) The systematic name is specified as the current position of the voice input person (step 3005) and output (step 300).
6). For example, in FIG. 2, the sensor (S2) 1003 is used to check the circulating water pump A and perform the inspection work of the T / D-RFP-A by the voice input person who is performing the inspection work along the route 1. When passing, sensor (S2) 1
The information that 003 is "installed between the circulating water system and the water supply system" and the route 1 determines from the information that "the water supply system is after the circulating water system", and outputs "water supply system". By the way,
As a method of knowing which route the voice input person is currently working on, the voice input person may register in advance before starting the inspection, or store the date and / or time information in the inspection route data. A method of automatically switching or reading the inspection route (step 300
4) At the time, inspection items such as a method of determining based on the date and / or time can be set and implemented.

FIG. 6 shows the structure of a voice recognition dictionary (partial dictionary) 1301 according to the first embodiment. Position detector 100
Since the system name is output from 1 as the position information, the partial dictionary is created for each system and is composed of a "system name" area and a "voice pattern" area. In the "voice pattern" area, "voice pattern data" used for recognition processing and "character string" corresponding to this voice pattern data are stored. Further, the pointer indicates the currently selected partial dictionary, and the voice recognition unit 1401 is provided to specify the selected partial dictionary.

FIG. 7 shows the processing of the partial dictionary selection unit 1201 according to the first embodiment. This process is performed by the position detection unit 100.
Executed when input from 1. Upon activation, the current position (system name) is fetched from the position detection unit 1001 (step 3101), and the partial dictionary corresponding to the system name indicating the current position is searched by referring to the system name area of the partial dictionary. Then (step 3103), the pointer is moved to the partial dictionary corresponding to the current position (step 31).
03).

FIG. 8 shows a voice recognition unit 1 according to the first embodiment.
The processing of 401 is shown. In this processing, when there is an input from the microphone 1801 (step 3201), the transmitting / receiving unit (mobile station)
Voice is transmitted from 1602 (step 3202), and voice is received by the transmitting / receiving unit (fixed station) 1601 (step 3).
203). The voice recognition unit 1401 inputs the received voice (step 3204) and performs a recognition process using the selected partial dictionary in the voice recognition dictionary (partial dictionary) storage unit 1301, that is, the partial dictionary on which the pointer is placed. Is performed (step 3205), and the recognition result is output to the computer 1501 (step 3206).

FIG. 9 shows inspection record data stored in the inspection record storage section 1502. The inspection record data stores the system name, inspection target, and inspection result. The system name and inspection target are stored in advance. The inspection result is stored in the position corresponding to the inspection system and the inspection target based on the character string output from the voice recognition unit 1401 by the computer 1501.

FIG. 10 shows a CRT 15 of the inspection report prepared based on the inspection record data in the inspection record storage section 1502.
The example displayed in 03 is shown. After inspection work CRT150
It is possible to display the inspection report on the screen 3, confirm the inspection contents on the CRT 1503, and output the inspection report to the printer 1504.

FIG. 11 shows a display example of the portable display 1701. The mobile display 1701 is used to confirm the inspection record while inputting voice, and the display data is provided from the computer 1501. The initial screen is in the state of waiting for input of the inspection target like the screen 1711. When the inspection target device "circulating water pump" is input by voice input, the "circulating water pump" is displayed as the inspection target as the screen 1712. In addition, the inspection items are displayed. The inspection item is provided as display data by referring to the inspection target data stored in advance in the inspection record storage unit 1502 by the computer 1501 based on the target device name input by voice. Next, when "Velocity" is input by voice to input "Velocity", the portion displaying the inspection contents of the velocity is highlighted (screen 1713), and the computer 1501 confirms that it is waiting for the input of "Velocity". Show. If you enter the speed value "70",
The value is displayed (screen 1714). Next, when "voice" is input by voice, the input position of "voice" is highlighted.
In this way, by entering the item and its value,
It is possible to import the inspection results as computer data. Also, if the value you entered is mistakenly recognized, or if you enter the value incorrectly,
It is possible to correct the value by re-inputting after uttering "correction". In the present embodiment, it has been described that the sensor is installed at the gate that isolates each system, and the partial dictionary is divided for each system name, but the sensor is installed near the device and the partial dictionary is provided for each device name. It is also possible to divide the dictionary and specify the device installed near the position where the voice input is performed.

As described above, in this embodiment, the sensor (installed between systems) installed in the plant is used to detect the passage of the voice input person, and the current position (system name) is specified. It is possible to switch and use partial dictionaries that have been divided in advance for each situation.Also, when the situation changes, that is, the inspection item or inspection route is changed, or the inspection time is abnormal, work conditions Even if it changes due to progress, etc., it will switch to a predivided partial dictionary by detecting the passage of a voice input person using a sensor and specifying the current position (system name), improving the recognition rate and The recognition time can be shortened. Further, by displaying the result of voice recognition on the mobile display, it is possible to reliably prevent erroneous input.

FIG. 12 shows a voice recognition device of a plant inspection system according to a second embodiment of the present invention. The voice recognition device includes a position detection unit 101 that detects the position of a voice input person.
1. The plant information storage unit 111 referenced by the position detection unit
1, partial dictionary selection unit 1211, voice recognition dictionary (partial dictionary) storage unit 1311, voice recognition unit 1411, computer 15
01, an inspection record storage unit 1502, a CRT 1503 for displaying the inspection record, a printer 1504 for outputting an inspection report based on the inspection record, a transmission / reception unit (fixed station) 1601 for transmitting data, a transmission / reception unit (mobile station) 1602, a mobile display 1701, and a microphone 1801.
This configuration is the same as that of the first embodiment except the position detection unit 1011, the plant information storage unit 1111, the partial dictionary selection unit 1211, the voice recognition dictionary (partial dictionary) storage unit 1311 and the voice recognition unit 1411. Hereinafter, a position detection unit 1011 and a plant information storage unit 1 different from the first embodiment
The 111, the partial dictionary selection unit 1211, the voice recognition dictionary (partial dictionary) storage unit 1311 and the voice recognition unit 1411 will be described.

FIG. 13 is a schematic diagram of a plant site according to the second embodiment. In FIG. 13, sensors (SD1 to SDn) 1012 to 1019 are installed on a route (route 1 and route 2) for a voice input person to inspect. The sensors (SD1 to SDn) 1012 to 1019 measure the distance from the sensor to the voice input person.
As a sensor for measuring the distance, for example, there is a technology such as a laser distance sensor by triangulation (Kondo et al .: “Digital signal processing in measurement / sensor”, pp.118-120, Shokoido, (1993)). .

FIG. 14 shows installation position data of the sensors (SD1 to SDn) 1012 to 1019 stored in advance in the plant information storage unit 1111 referenced by the position detection unit 1011. In the sensor installation position data, the installation position coordinates (Xs, Ys) of the sensor at the plant site are set. It also stores in which direction the sensor is attached with respect to the X and Y axes. This is the △ in Figure 14
It corresponds to x and Δy. In FIG. 14, + indicates that the sensor measures the distance in the direction in which the x or y coordinate increases, and-in the direction in which the coordinate decreases. For example, the sensor (SD
If the distance to the voice input person detected by 1) is L1,
Since the sensor (SD1) measures the distance in the increasing direction of x, the position of the voice input person is (SD1x + L
1, SD1y).

FIG. 15 shows equipment placement position data stored in advance in the plant information storage unit 1111 referenced by the position detection unit 1011. The equipment position data also sets the installation position by coordinates (Xm, Ym) with respect to the X and Y axes of the same plant site as the sensor.

FIG. 16 shows the processing of the position detector 1011 according to the second embodiment. In this processing, when the sensor detects a voice input person (step 4001), the distance from the sensor detecting the voice input person to the voice input person is fetched (step 4002), and the set position data of the sensor detecting the voice input person is acquired. It is read from the plant information storage unit 1111 (step 4003), and the position coordinate of the voice input person is calculated (step 4004). Next, a device near the position coordinate of the voice input person is searched with reference to the device installation position data in the plant information storage unit 1111 (step 400).
5). If there is only one search result (step 400)
6) The search result is output and the processing is terminated (step 40
08), if there are a plurality of them, they are rearranged in the order of decreasing distance from the position coordinates (step 4007) and output (step 4008).

FIG. 17 shows the structure of the voice recognition dictionary 1311 according to the second embodiment, and FIG. 18 shows the structure of the partial dictionary 1311. Since the device name is output from the position detection unit 1011 in the voice recognition dictionary, a partial dictionary is created for each plant / system / device to have a hierarchical structure of systems and devices.
For example, the partial dictionary of the plant is a partial dictionary showing the vocabulary related to the plant, the partial dictionary of the water supply system is a partial dictionary showing the vocabulary related to the lower system of the plant, the water supply system, and N36 valve. The partial dictionary of is a partial dictionary showing the vocabulary related to the N36 valve, which is a subordinate device of the water supply system. The oil filter and the partial dictionary of T / D-RFP-A have a sibling relationship of N36 valves. Each partial dictionary is
It consists of a "device" or "system name" or "plant" area and a "voice pattern" area. In the voice pattern area, "voice pattern data" and "character string" corresponding to this voice pattern data are stored. It is stored.

FIG. 19 illustrates an example of prioritizing partial dictionaries, which is performed by the partial dictionary selection processing of the partial dictionary selection unit 1211 according to the second embodiment. In FIG. 19, M1 and M2 are device names output by the position detection unit 1011 and M1 is the device closest to the voice input person. Further, P1 to P4 are priorities, P1 has a high priority, and P2 and P3 have a low priority. This priority is
When the voice recognition unit 1411 performs the recognition process of the input voice,
This shows the order of the partial dictionary 1311 to be referred to.

FIG. 20 is a partial dictionary 1 according to the second embodiment.
311 shows the process of prioritizing 311. The flow of processing will be described using the example of FIG. As shown in (1) of FIG. 19, the position detection unit 1011 displays the device name “N36 valve” (M1).
If “NO” is output, the device name is “N36 valve” (M1) only (step 4101).
The priority P1 is set to the partial dictionary for 1) (step 4110), and the priority P2 is set to the partial dictionary corresponding to the "water supply system" one level higher (step 4111). Next, the "valve LPS" in the same floor as the "N36 valve" (M1)
The priorities P3 to Pk are randomly set in the partial dictionary corresponding to "V""T / D-RFP-A" (step 411).
2). Next, as shown in (2) of FIG.
11 is the device name "N36 valve" (M1) and "T / D-RFP"
-A "(M2) is output, the device name is plural, and" N36 valve "(M1) and" T / D-RFP-A ".
(M2) belongs to the same layer (step 410).
1-4102), the priority P1 is set in the partial dictionary for "N36 valve" (M1), and the priority P2 is set in the partial dictionary for "T / D-RFP-A" (M2) (step 412).
1), "N36 valve" (M1) and "T / D-RFP-A"
(M2) The priority P3 is set in the partial dictionary corresponding to the "water supply system" one level higher. Then, "N36 valve" (M
1) and "T / D-RFP-A" (M2), the priorities P4 to Pn are randomly set in the partial dictionary corresponding to "valve LPSV" in the same hierarchy. Next, as shown in (3) of FIG. 19, the position detection unit 1011 causes the device name “valve LPSV” (M
1) and "hydrogen gas pressure gauge" (M2) are output,
Since there are multiple device names, and the "valve LPSV" (M1) and the "hydrogen gas pressure gauge" (M2) do not belong to the same layer (steps 4101 to 4102), first, the "valve LPSV"
The partial dictionary corresponding to "V" (M1) has priority P1, and the partial dictionary corresponding to "hydrogen gas pressure gauge" (M2) has priority P2.
Is set (step 4130), and the "valve LPSV" (M
1) Priority P3 is assigned to the partial dictionary corresponding to "water supply system" which is one level higher, and priority P4 is assigned to the partial dictionary corresponding to "hydrogen gas pressure gauge" (M2) which is one level higher than "internal power supply system". It is set (step 4131).

FIG. 21 shows the processing of the voice recognition unit 1411 according to the second embodiment. In this processing, when there is an input from the microphone 1801 (step 4201), a voice is transmitted from the transmitting / receiving unit (mobile station) 1602 (step 4202),
The transmitter / receiver (fixed station) 1601 receives voice (step 4203). The voice recognition unit 1411 inputs the received voice (step 4204), refers to the voice recognition dictionary 1311, and selects partial dictionaries in order of priority (step 4).
205) and after searching for a voice pattern in the partial dictionary that matches the voice pattern of the input voice (step 420)
6) If the search is successful (step 4207), the character string corresponding to the matched voice pattern is output to the computer 1501 (step 4208), and if the search is not successful (step 4207), the recognition is unrecognizable. Calculator 1
It is output to 501 (step 4209).

As described above, in this embodiment, the sensor installed in the plant (a plurality of sensors are installed on the voice input person's path) is used to determine the position of the voice input person by the X and Y coordinates provided in the plant. By detecting, the device near the current position is specified. Further, based on this, as a function hierarchy of the system / device, priorities are given to the partial dictionaries created in advance, and in the voice recognition processing, the partial dictionaries divided by the function hierarchy are switched based on this priority. As a result, efficient voice recognition can be performed by using the finely divided partial dictionary, and the recognition rate can be improved and the recognition time can be shortened. In addition, in the present embodiment, the system / equipment of the plant is expressed as a function hierarchy, and the partial dictionary is prepared accordingly, so that it is possible to efficiently organize frequently used words during plant maintenance. Thereby, the construction of the partial dictionary can be facilitated, and selection of the partial dictionary and voice recognition at the time of voice input can be made efficient.

FIG. 22 shows a voice recognition device of a car navigation system according to the third embodiment of the present invention. The voice recognition device includes a position detection unit 1021 and a dictionary selection unit 122.
1, voice recognition dictionary storage unit 1321, voice recognition unit 142
1, calculator 1521, display unit 1722, microphone 180
1. It is composed of map data 1921 in which graphic data of a map showing a position of a road or the like corresponding to latitude and longitude and a prefecture / city group name are stored. The position detection unit 1021 measures the latitude and longitude of the current position using a global aviation positioning system (hereinafter abbreviated as GPS). GPS is a system developed by the United States of America, which simultaneously receives radio waves from three satellites of a satellite group launched into an orbit around the earth and measures a two-dimensional position. As a familiar item, it is used to display the current position on a map displayed on a display device in a car navigation system. By the way, if radio waves are simultaneously received from four satellites,
Since it can measure three-dimensional position including altitude, it is also used in aircraft autopilots. Dictionary selection unit 122
1 is the position coordinates from the position detection unit 1021, and the computer 15
The voice recognition dictionary storage unit 1321 is selected based on the destination prefecture name and the map data obtained from 21. Further, the dictionary is switched by a signal from the voice recognition device 1421. The voice recognition unit 1421 performs a recognition process on the voice input from the microphone 1801 using the dictionary in the voice recognition dictionary storage unit 1321 selected by the dictionary selection unit 1221. When there is no corresponding word in the dictionary selected by the dictionary selection unit 1221, a dictionary switching signal is output to the dictionary selection unit 1221. The computer 1521 has a voice recognition unit 1
Based on the recognition result from 421, processing such as output of the target position to the dictionary selection unit 1221, display of map data on the display unit 1722, and search for a route from the current position to the target position is performed.

FIG. 23 shows a display example of the map information on the display unit 1722. The position P in FIG. 23 is the current position, and the latitude and longitude are detected by the position detection unit 1021 and displayed on the map.

FIG. 24 shows the structure of the voice recognition dictionary 1321 according to the third embodiment. The voice recognition dictionary 1321 includes one common dictionary and a plurality of city group name-based partial dictionaries, and the common dictionary is referred to in each recognition process. In the common dictionary (see FIG. 29), "voice pattern data" and "character string" corresponding to this voice pattern data
Is stored. Further, the partial dictionary by city group name is composed of an area for storing "city group name" and an area for storing "voice pattern". In the voice pattern area, "voice pattern data" and this voice pattern are stored. The "character string" corresponding to the data is stored.

FIG. 25 shows the processing of the dictionary selection unit 1221 according to the third embodiment. The dictionary selection unit 1221 inputs the latitude and longitude of the current position P from the position detection unit 1021 (step 5001), and refers to the map data 1921 to identify the city group name of the current position P (step 5002). Next, the city group name column of the city group name-specific partial dictionary is sequentially referenced to search for the city group name-specific partial dictionary that matches the specified city group name, and the pointer (see FIG. 24) is set (step 500).
3).

FIG. 26 shows the processing of the voice recognition unit 1421 according to the third embodiment. When the voice recognition unit 1421 receives an input from the microphone 1801 (step 5101), the voice recognition unit 1421 inputs a voice (step 5102),
After referring to the voice recognition dictionary storage unit 1321, selecting a partial dictionary with a pointer (step 5103), searching for a voice pattern of the partial dictionary that matches the voice pattern of the input voice (step 5104), and then matching The character string corresponding to the selected voice pattern is output to the computer 1521 (step 5105).

As described above, in the present embodiment, when the name of a town or village is input by voice, the name of the city group is specified in advance based on the current position measured (latitude and longitude) by the position detection unit, so that the city group can be identified in advance. It is possible to switch and use partial dictionaries created by group name, and the time required to reach the destination changes due to changes in the situation, that is, changes in the traveling route, or due to congestion on the traveling route. Even if
By specifying the city group name based on the current position, it is possible to switch the partial dictionaries divided in advance by city group name. In this way, it is possible to efficiently select a partial dictionary containing words that are highly likely to be uttered according to the situation of the voice input person, so that the recognition rate is improved and the recognition time is shortened even in the voice recognition device of the car navigation system. Can be achieved.

FIG. 27 shows a voice recognition device of a car navigation system according to the fourth embodiment of the present invention. The voice recognition device includes a position detection unit 1021 and a dictionary selection unit 123.
1, voice recognition dictionary storage unit 1331, voice recognition unit 143
1, calculator 1531, display unit 1722, microphone 180
1. It is composed of map data 1921 in which graphic data of a map showing a position of a road or the like corresponding to latitude and longitude and a prefecture / city group name are stored. This configuration is the same as the third embodiment except for the dictionary selection unit 1231, the voice recognition dictionary storage unit 1331, the voice recognition unit 1431, and the computer 1531. The voice recognition unit 1431 performs a recognition process on the voice input from the microphone 1801 using the dictionary in the voice recognition dictionary storage unit 1331 selected by the dictionary selection unit 1231. The computer 1531 performs processing such as output of the target position to the dictionary selection unit 1231, display of map data on the display unit 1722, and search for a route from the current position to the target position, based on the recognition result from the voice recognition unit 1431. I do.

FIG. 28 shows a display example of the map information on the display unit 1722. The position P in FIG. 28 is the current position, and the latitude and longitude are detected by the position detection unit 1021 and displayed on the map.

FIG. 29 shows the structure of a voice recognition dictionary 1331 according to the fourth embodiment. The voice recognition dictionary 1331 includes one common dictionary and a plurality of city group name-based partial dictionaries. The common dictionary includes an area for storing “priority” indicating the order in which the voice recognition unit 1431 refers to the dictionary and an area for storing “voice pattern”. In the voice pattern area, "voice pattern data" and "character string" corresponding to this voice pattern data are stored.
The city group name-based partial dictionary is composed of an area for storing “priority”, an area for storing “city group name”, and an area for storing “voice pattern”. The "voice pattern data" and the "character string" corresponding to this voice pattern data are stored.

FIG. 30 shows the processing of the dictionary selection unit 1231 according to the fourth embodiment. The dictionary selection unit 1231 sets a priority of 1 in the common dictionary (step 6001), and inputs the latitude and longitude of the current position P from the position detection unit 1021 (step 6002). Next, with reference to the map data 1921, the city group name of the current position P is specified (step 600).
3) The city group name column of the city group name-specific partial dictionary is sequentially referred to, the city group name-specific partial dictionary that matches the specified city group name is searched, and priority 2 is set (step 6004). Next, the destination (prefecture name) is input from the computer 1531, the prefectural border P'with the closest destination (prefecture name) from the current position P is obtained (step 6005), and the straight line PP 'passes. City group names are sequentially searched from the current position P side toward P ', and the priority (from priority 3) is set in the partial dictionary corresponding to the searched city group name (step 6006). In the dictionary selection process, when the destination (prefecture name) is not input to the computer 1531, steps 6005 to 6006 are not performed.

FIG. 31 shows the dictionary selecting section 123 at the current position P in the case of the destination (prefecture name) in the map shown in FIG.
1 shows the result of prioritizing the dictionary. The common dictionary, which is a dictionary that is referred to each time, has a priority of 1, the partial dictionary corresponding to the city group name at the current position P has a priority of 2, a city of M prefectures, a Y city of M prefectures, and a M city of Y.
Prefecture Z city is set with priority 3 and 4, respectively.

FIG. 32 shows the processing of the voice recognition unit 1431 according to the fourth embodiment. When the voice recognition unit 1431 receives an input from the microphone 1801 (step 6101), the voice recognition unit 1431 inputs a voice (step 6102),
After referring to the voice recognition dictionary storage unit 1331, the partial dictionaries are selected in order of priority (step 6103), the voice pattern of the partial dictionary that matches the voice pattern of the input voice is searched (step 6104), and then the voice patterns match. The character string corresponding to the voice pattern is output to the computer 1531 (step 6105).

As described above, in this embodiment, when the name of a town or village is input by voice, based on the current position measured (latitude and longitude) by the position detection unit and / or based on the destination and the current position, It is possible to give priority to partial dictionaries created in advance for each city group name, and switch the partial dictionaries based on this priority in the voice recognition process. In this way, it is possible to efficiently select a partial dictionary containing words that are highly likely to be uttered according to the situation of the voice input person. Therefore, even in the voice recognition device of the car navigation system, the recognition rate is improved and the recognition time is shortened. Can be achieved.

[0042]

As described above, according to the present invention,
The voice recognition dictionary is divided into partial dictionaries and stored in advance.The position of the voice input person or car is detected, and the partial dictionary is selected based on the position of the voice input person or car. Since the voice recognition process is performed by using it, the recognition rate of the voice recognition device can be improved and the recognition time can be significantly shortened as compared with the case where all words are registered in one dictionary. Also, by taking in the position of the voice input person or the car, and adopting the priority to dynamically switch the partial dictionaries, it is possible to efficiently switch the partial dictionaries even when various situations change, The recognition rate of the voice recognition device can be improved and the recognition time can be greatly shortened. In addition, the system / equipment of the plant is expressed as a function hierarchy, and partial dictionaries are prepared in accordance with this, so it is possible to efficiently organize frequently used words during plant maintenance. The construction can be facilitated, and the selection of the partial dictionary and the voice recognition at the time of voice input can be made efficient. Further, by using the voice recognition device according to the present invention for a plant inspection system and a car navigation system, the efficiency of inspection and navigation can be improved respectively, and thus the safety of the plant and the maneuverability of the vehicle can be improved. .

[Brief description of drawings]

FIG. 1 is a configuration diagram of a voice recognition device of a plant inspection system according to a first embodiment of the present invention.

FIG. 2 is a schematic diagram of a plant site according to the first embodiment.

FIG. 3 is sensor installation position data according to the first embodiment.

FIG. 4 Inspection route data.

FIG. 5 is a process of the position detection unit according to the first embodiment.

FIG. 6 is a voice recognition dictionary (partial dictionary) according to the first embodiment.
Configuration diagram of.

FIG. 7 is a process of a partial dictionary selection unit according to the first embodiment.

FIG. 8 is a voice recognition process according to the first embodiment.

FIG. 9 is an inspection record storage unit.

FIG. 10 is a display example of a CRT.

FIG. 11 is a display example of a mobile display.

FIG. 12 is a configuration diagram of a voice recognition device of a plant inspection system according to a second embodiment of the present invention.

FIG. 13 is a schematic diagram of a plant site according to a second embodiment.

FIG. 14 is sensor installation position data according to the second embodiment.

FIG. 15 is device installation position data according to the second embodiment.

FIG. 16 is a process of the position detection unit according to the second embodiment.

FIG. 17 is a configuration of a voice recognition dictionary according to a second embodiment.

FIG. 18 is a configuration of a partial dictionary according to the second embodiment.

FIG. 19 is an example of prioritizing partial dictionaries according to the second embodiment.

FIG. 20 is a process of prioritizing partial dictionaries according to the second embodiment.

FIG. 21 is a voice recognition process according to the second embodiment.

FIG. 22 is a configuration diagram of a voice recognition device of a car navigation system according to a third embodiment.

FIG. 23 is a display example of map information according to the third embodiment.

FIG. 24 is a configuration of a voice recognition dictionary according to a third embodiment.

FIG. 25 is a process of the dictionary selection unit according to the third embodiment.

FIG. 26 is a voice recognition process according to the third embodiment.

FIG. 27 is a configuration diagram of a voice recognition device of a car navigation system according to a fourth embodiment.

FIG. 28 is a display example of map information according to the fourth embodiment.

FIG. 29 is a configuration of a voice recognition dictionary according to the fourth embodiment.

FIG. 30 is a process of the dictionary selection unit according to the fourth embodiment.

FIG. 31 is an example of prioritizing a dictionary according to the fourth embodiment.

FIG. 32 is a voice recognition process according to the fourth embodiment.

[Explanation of symbols]

1001, 1011, 1021 Position detection unit 1002-1006 Sensor for detecting passage of voice recognizer 1012-1019 Distance sensor 1101, 1111 Plant information storage unit 1201, 1211 Partial dictionary selection unit 1221, 1231 Dictionary selection unit 1301, 1311, 1321 , 1331 voice recognition dictionary (partial dictionary) storage unit 1401, 1411, 1421, 1431 voice recognition unit 1501, 1521, 1531 computer 1502 inspection record storage unit 1503 CRT 1504 printer 1601 transceiver unit (fixed station) 1602 transceiver unit (mobile station) 1701 mobile display 1711 to 1715 display screen example of mobile display 1722 display unit 1801 microphone 1921 map data

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G01D 21/00 G01D 21/00 Q G01S 5/02 G01S 5/02 Z // G01C 21/00 G01C 21/00 H

Claims (10)

[Claims] 1. A voice recognition device that divides a voice pattern dictionary into partial dictionaries in advance and uses the partial dictionaries by switching the partial dictionaries to recognize an input voice, wherein the portion to be used is based on the position of a voice input person. A voice recognition device characterized by selecting a dictionary, comparing the input voice pattern with the partial dictionary, and recognizing the input word. 2. A voice recognition device that divides a voice pattern dictionary into partial dictionaries in advance and switches the partial dictionaries to recognize an input voice, and position detection means for capturing the position of a voice input person, and the voice input. The partial dictionary selecting means for switching the partial dictionary based on the position of the person, and comparing the selected partial dictionary with the input voice pattern,
A voice recognition device comprising voice recognition means for recognizing an input word. 3. The on-site inspection system for a plant according to claim 1 or 2, wherein the partial dictionary of voice patterns comprises position data of a voice input person and plant inspection route data and / or inspection item data. The system and / or equipment covered by the plant inspection has been identified,
A voice recognition device, which is switched based on the specified system and / or device to be inspected. 4. The voice recognition according to claim 3, wherein the position data of the voice input person is obtained from a detection means installed near the gate between each system forming the plant site and / or the plant equipment. apparatus. 5. The plant inspection system according to claim 1 or 2, wherein the partial dictionary of voice patterns is a target of plant inspection based on position data of a voice input person and arrangement data of plant equipment. A voice recognition device characterized in that a system and / or a device in use is specified, and switching is performed based on the specified system and / or device to be inspected. 6. The plant according to claim 5, wherein the position coordinates are set at the plant site, and the position data of the voice input person is obtained from the coordinates of the installation position of the detection means installed at an arbitrary position on the plant inspection route, The arrangement data is obtained from the coordinates of the installation position where the device is installed. 7. The voice recognition device according to claim 5, wherein the partial dictionary of voice patterns is created by expressing a system and / or equipment of a plant as a functional hierarchy and / or giving priority to the system. . 8. A voice recognition device that divides a voice pattern dictionary into partial dictionaries in advance and switches the partial dictionaries to recognize an input voice, which is a car navigation system and detects a position of a current position of a vehicle. Means and the partial dictionary selecting means for switching the partial dictionary while referring to the map data based on the current position of the vehicle, and the selected partial dictionary and the input voice pattern are compared, and the input word is input. A voice recognition device having voice recognition means for recognizing. 9. A voice recognition device for dividing a voice pattern dictionary into partial dictionaries in advance and switching the partial dictionaries to recognize an input voice, which is a car navigation system and detects a position of a current position of a vehicle. Means, based on the current position of the vehicle, while referring to map data, sets a priority in the partial dictionary up to the destination, and selects partial dictionary selecting means for switching the partial dictionary, and in order of priority. A voice recognition device comprising voice recognition means for recognizing an input word by comparing a selected partial dictionary with an input voice pattern. 10. The voice recognition device according to claim 9, wherein the voice pattern dictionary includes one common dictionary and a plurality of partial dictionaries, and the highest priority is set in the common dictionary.